Data Bench Documentation

Welcome to the Data Bench docs. This guide walks you through installing the CLI, running your first benchmark, and understanding the results — while keeping the navigation bar and page layout consistent with the rest of the site.

Tip: Use the sidebar to navigate to dedicated pages for installation, CLI usage, and your first steps.

Install dependencies

Data Bench requires git and Docker.

# macOS
brew install git --quiet
brew install --cask docker

Install the CLI

You can install the CLI using your preferred Python tool.

# uv
uv tool install terminal-bench

# pip
pip install terminal-bench

After installation, run tb --help to view commands and options.

tb --help

Common commands

  • tb run — run the benchmark harness
  • tb tasks — inspect available tasks
  • tb datasets — manage datasets
  • tb runs — list or inspect previous runs

First steps

# 1) Create a new run
tb run --model my-model

# 2) Open the dashboard
tb runs open latest

Need more? Explore tasks and examples, or contribute new datasets. If you have questions, open an issue or reach out on Discord.